A Flexible Random Forest Regressor-Based Algorithm for Predicting Heart Rate Through Facial Video

Abstract

A continuous, non-invasive monitoring of vital signs like heart rate is crucial in modern healthcare for early detection of cardiovascular abnormalities and supporting telemedicine growth. Traditional methods, such as ECG and PPG, rely on contact-based sensors that are often uncomfortable, costly, and unsuitable for long-term or remote monitoring. This project introduces a fully software-based, contactless heart rate estimation system utilizing a standard webcam. The system applies computer vision techniques to detect facial regions of interest (forehead and cheeks) and machine learning models to interpret subtle blood-flow-induced color variations from facial video. The proposed pipeline includes face detection, region tracking, signal extraction, preprocessing, feature computation, and machine learning prediction using Random Forest Regressor Model. The developed system performs reliably across different lighting conditions, skin tones, and facial movements, achieving near real-time operation. This work highlights the potential of using accessible consumer hardware for accurate, contactless physiological monitoring, paving the way for integration in telemedicine, remote patient monitoring, and personal fitness applications.